|
Historically, programmers have attempted to test artificial intelligence limits by creating chess-playing robots with the ability to analyze thousands of moves in advance. The development of these computers led to the famous match-up between IBM's Deep Blue and chess champion Gary Kasparov in 1996.
However, as challenging a game as chess is for computers, poker is more complicated. This stems from the fact that chess players, artificial or real, can see all the pieces on the board, all the time. Poker players, however, have to deal with hidden details, such as which cards will show on the flop and which cards your opponent has been dealt.
Tuomas Sandholm, a professor at Carnegie Mellon University, specializes in the important connection between computer intricacy and strategy. He runs his own company, named CombineNet, which develops algorithms to optimize procurement for its clients. He developed a poker-playing technology that attempts to navigate the millions of different poker hands possible in a game of Texas Hold'Em. The technology, referred to as GS2, was put to the test this past July when Sandholm entered it into a Boston Tournament sponsored by the American Association for Artificial Intelligence.
There are about 26 million possibilities for poker hands in just the second round of a Texas Hold'Em hand. That is too many for a computer to process. The GS2 deals with this by using an algorithm that reduces the number of hands to 2,465, grouping them by strategic similarity. The computer has yet to beat a top human poker player.
"The research problem is how to come up automatically with better and better strategies," Sandholm says. "The computational and combinatorial complexity of solving the game makes this enormously challenging."
|
